AI to boost self-regulated learning

Learning to learn is important if you want to do well academically and while self-regulation is tricky, AI might be able to help.

Software made by an international group of universities – The FLoRA project – will help to track students’ learning habits, identifying areas that need work and generating learning scaffolds for each towards improving their self-regulation around learning.

Personalised scaffolds will be offered as an extension of two common educational technologies that are freely available and open source – Moodle and Hypothes.is.

“In essence, existing resources that are used by learners can still be used. Except that we are creating an extra layer of AI that can enhance learning experience and performance,” says Professor Dragan Gasevic, Director of the Centre for Learning Analytics at Monash and a Chief Investigator in the project.

The software capitalises on the multitude of digital traces left when students use learning software. Data signatures discovered with machine learning form a foundation for personalised scaffolding. As each learner will have a different data signature, custom tailored scaffolds will be offered to each learner. Data signatures inform when, what, and how many scaffolds are offered to the learners. The actual content (the what part) of personalised scaffolds is fully informed by existing research in educational psychology and the learning sciences.

“Specifically, we use machine learning algorithms to analyse these digital traces to help us identify common ‘data signatures’ that are unique for each learner. Data signatures capture, in fact, common study patterns that learners follow. Data signatures can represent, for example, that learners use relatively ineffective learning strategies. Or, learners do not sufficiently check their own knowledge gained while previously learning.”

He says that one of the key imperatives of modern education is to prepare lifelong learners who can learn independently. Learners need to have skills for self-regulated learning.

“This implies that learners need to have a capacity of ‘learning to learn’. Prior research has shown high skills for self-regulated learning lead to improved learning performance.

“However, learners often experience difficulties self-regulating their learning adequately. This project is precisely focused on addressing these limitations and helping learners enhance their skills for self-regulated learning.”

Scaffolds or teaching guides will mostly be presented in the form of digital prompts that are offered to the learners. The prompts are delivered as dialogue boxes in the user interface of the digital learning environment the researchers are working on in the project.

“The main point of scaffolds is to offer suggestions or guidance that correspond to learners’ needs as revealed by their digital signatures. For example, learners are reminded to revise certain information, to engage with the information where learning objectives are provided, or to reassess their learning goals. 

“An equally critical dimension of the process of creating scaffolds is the emphasis on human factors. That requires learners to be involved in the co-design of scaffolds to make sure learners can understand the value of scaffolds.”

“An alternative and a bit less intrusive approach to offering scaffolds we are working on is to send messages to the learners and only highlight their receipt in the user interface of the digital learning environment. In this alternative case, the learners have more autonomy in making decisions on when and how they will engage with the scaffolds,” he says.

Once the software is up and running Gasevic expects higher learning gains for the learners who receive personalised scaffolds in comparison to the learners who receive general scaffolds.

“General scaffolds are often used in education; they are based on the recommendations from the literature that make assumptions about learners on average. However, there is no such thing as a learner on average. As such, general scaffolds work well for some learners, but can be detrimental for others. With the use of digital signatures, we will be able to personalised scaffolds to cater to individual needs of each learner.

“As done with the technologies we are extending (Moodle and Hypothes.is), our plan is to share the results of our project as freely available and open source technologies. We will also offer resources that will demonstrate best practices and initiate a community of users and developers.

The technology has been featured in the prestigious EDUCAUSE Horizon Report and Prof Gasevic explains that it is a huge accolade for the Monash team.

“The EDUCAUSE annual report is the most influential publication regarding the use of technology in higher education and it has an even greater impact on the policy and strategic planning in almost every higher education institution around the world.

“To have our project recognised by this publication is a fantastic achievement and one that showcases the increasing influence technology has on education,” said Prof Gasevic.

There is a trove of data on students which is often not used to its full extent, but Prof Gasevic says there is a paucity of information around self-regulation during study.

“It is true that lots of data about learners is collected. However, little data about how learners self-regulate their learning is actually collected. Learning is a very dynamic process and we need to collect much more granular data than presently captured by digital technologies. We address this limitation by adding additional (instrumentation) tools into digital learning environments – e.g., tools for planning study time, highlighting content, or setting goals. Not only do these tools help us increase the quality and granularity of data, but they also assist learners to better regulate their learning. Machine learning without high quality and granular data is not particularly helpful in enhancing learning.”

The project is of four years’ duration and the group is at the start of the third year. In addition to the work on technologies, the research involves four laboratory studies with learners and the final field study that will involve learners from four different countries (Australia, Germany, The Netherlands, and United Kingdom). The University of Edinburgh, Technical University of Munich, Radboud University and Monash University are partners in FLoRA.